th

The 14 World Conference on Earthquake Engineering

October 12-17, 2008, Beijing, China

SCALING OF EARTHQUAKE ACCELEROGRAMS FOR NON-LINEAR DYNAMIC ANALYSES

TO MATCH THE EARTHQUAKE DESIGN SPECTRA
1 2
Y. Fahjan and Z. Ozdemir
1
Asst. Prof., Dept. of Earthquake and Structural Science, Gebze Institute of Technology, Gebze, Kocaeli,
Turkey
2
Ph.D. Candidate, Dept. of Earthquake Engineering, Bogazici University, Istanbul, Turkey
Email: fahjan@gyte.edu.tr, zuhal.ozdemir@boun.edu.tr

ABSTRACT :

A complete seismic design of structures requires linear and nonlinear time-history analyses especially for
special type buildings. Seismic design codes generally define ground shaking in the form of a response
spectrum of acceleration and allow using response spectrum compatible time history records in linear and
nonlinear time history analyses. These records can be obtained from natural earthquake records, or can be
generated synthetically and artificially. Although using real earthquake records has many advantages, there may
exist lack of strong motion earthquake records to satisfy seismological and geological conditions and
requirements defined in seismic codes. Artificial accelerograms whose response spectra closely compatible to
design response spectra can be generated in either time or frequency domain. Matching techniques are based on
scaling of the selected time history in time domain; filtering actual motion in frequency domain by its spectral
ratio with the design target spectrum; or elementary wavelets are added or subtracted from the real time history
to match a target design spectrum. In this study, the spectrum matching procedures for real accelerograms are
summarized and applied to selected real acceleration records to match the proposed Type 1 elastic design
spectrum given in the Eurocode 8 for specified seismic region and soil type. Artificial accelerograms, which are
compatible with the selected design spectrum, are generated according to specified scenario earthquake. The
linear and nonlinear response of single degree of freedom system subjected to the modified and artificially
generated time histories acceleration records are compared and the advantages and disadvantages of each one
are discussed.

KEYWORDS: Matching techniques, earthquake records, spectrum compatible, artificial records,

non-linear analysis

1. INTRODUCTION

Seismic design is traditionally performed for most common structures by the means of equivalent lateral static
loading or modal spectrum analyses. Nevertheless, in some cases such as, irregular, highly ductile, critical or
higher modes induced structures, conventional response spectral analyses are not capable of estimating maximum
responses of linear systems, for which a time-integration scheme is deemed more appropriate (Preumont, 1984).
Seismic design codes generally define ground shaking in the form of a response spectrum of acceleration and
permit to use spectrally matched natural accelerograms recorded during earthquakes, spectrum compatible
artificially generated and synthetic ground motions for the linear or nonlinear analysis of structures in linear and
nonlinear time domain analyses.

The attraction of using natural accelerograms is due to the increase of available strong ground motion databases.
In the case of using real earthquake records, they have to be organized to match the design spectrum throughout
the full spectral range or only over a portion of a specified range that is of interest to the design. Spectral matching
may be performed in either the time domain or the frequency domain in such ways: The spectral acceleration
values of the selected time history are simply scaled up or down uniformly; an actual motion is filtered in
frequency domain by its spectral ratio with the design target spectrum; elementary wavelets are added or
 th
The 14 World Conference on Earthquake Engineering
October 12-17, 2008, Beijing, China

subtracted from the real time history to match a target design spectrum (RSPMatch). The spectrum matching
procedure begins with an acceleration time history whose characteristics reasonably represent the ground motions
expected for the site. Therefore, selection and matching of the input ground motions are critical issues to perform
linear and nonlinear response history analyses. The drawback of the use of real time histories lies in trying to
match a single ground motion to a design response spectrum that is not intended to represent the motion from an
individual earthquake (Naeim and Kelly, 1999). The design response spectrum is generally a result of a statistical
analysis that considers the influence of several seismic sources simultaneously, whence the response at different
periods may be driven by earthquakes in different sources and the spectrum is effectively the envelope of spectra
corresponding to scenarios in each of the sources (Reiter, 1990).

Despite the continued growth of the global strong ground motion databank, suitable records, which have the
same earthquake parameters with the site considered such as magnitude, rupture mechanism, source-to-site
distance and site classification, can not be obtained in some circumstances (Bommer et al., 2003). This can
impose the need to generate spectrum compatible artificial earthquake records using spectral density function
and random phases (e.g., SIMQKE) or using non-stationary stochastic vector processes (e.g., TARSCTHS) or
synthetically simulated records.

In this study, real earthquake records are selected simply based on magnitude, fault type and ground types
prescribed in the Eurocode 8 (2003). The spectrum matching procedures for real accelerograms are summarized
and applied to selected real acceleration records to match the proposed Type 1 elastic design spectrum given in
the Eurocode 8 for specified seismic region and soil type. Artificial accelerograms, which are compatible with
the selected design spectrum, are generated according to specified scenario earthquake. The resulting time
histories are investigated in terms of suitability as input to time history analyses of civil engineering structures.
The linear and nonlinear responses of single degree of freedom system subjected to these acceleration records
are obtained and the advantages and disadvantages of each one are discussed.

2. SELECTION OF REAL EARTHQUAKE RECORDS FOR SPECTRUM MATCHING

In seismic design codes, seismic scenario, which is based on a pair of magnitude, distance and soil conditions,
is generally represented by means of a spectral target shape. Guidance given in seismic design codes on how to
select appropriate real records is usually focused on compatibility with this response spectrum rather than
seismological parameters. Therefore, real earthquake records, which have similar characteristics (magnitude,
distance, site condition and faulting type) with the site under consideration, have to be selected to match elastic
response spectrum given in the code. When selecting the earthquake records, it is desirable to use earthquake
magnitudes within 0.25 magnitude units of the target magnitude (Stewart et al, 2001). Selection of records
having appropriate fault-site distances is important especially for near-fault sites. Site conditions have a major
effect on the characteristics and frequency content of the strong ground motion records. Even though the ground
motions are amplified in soft soils, the high frequency motions are attenuated. Also, in order to preserve
non-stationary characteristics of the initial time history, it is essential to start with an acceleration time history
whose spectrum is as close to the target spectrum as possible in the period range of interest. A close initial fit
also ensures a speedy convergence to the design values.

3. SPECTRAL MATCHING METHODS

There are three methods for modifying actual time histories and two methods to artificially generate time
histories to match a given design spectrum:

3.1. Ground Motion Scaling in Time Domain

In this approach, recorded motion is simply scaled up or down by a constant scaling factor uniformly to find out
 th
The 14 World Conference on Earthquake Engineering
October 12-17, 2008, Beijing, China

the best match to the target spectrum over a period range of interest, without changing the frequency content. It
could be stated that the accelerograms should only be scaled in terms of amplitude. The procedure is based on
minimizing the differences between the scaled motion’s response spectrum and target spectrum in a least-square
sense. The methodology proposed herein considers as “Difference” the squared scaled-to-target difference,
evaluated by the integral (Nikolaou, 1998),

[α S ] dT
TB 2
actual
Difference = ∫ a (T) − Sat arg et (T) (4.1)
TA
where, Satarget and Saactual are target acceleration response spectrum and acceleration spectrum of the given
(actual) time history, respectively. α, T, TA and TB are is scaling factor, period of oscillator, lower period of
scaling, and upper period of scaling, respectively. In order to minimize the difference, the first derivative of the
difference function with respect to the scaling factor has to be zero which lead to definition of scaling factor α,
in a discrete form, as
TB
actual
∑ (Sa (T) Satar get (T ) )
d Difference T = TA
min Difference ⇒ =0⇒ α= TB
(4.2)
dα actual 2
∑ (Sa (T ) )
T = TA

3.2. Spectral Matching in Frequency Domain

A frequency domain matching methodology uses an actual record to produce a similar motion that matches
almost perfectly a target (design) spectrum. In this method, an actual motion is filtered in frequency domain by
its spectral ratio with the design target spectrum. Fourier spectral amplitudes of an input motion are modified
while the Fourier phases of that remain unchanged during the entire procedure. Preservation of phase
characteristics is important for non-linear time domain analyses, because the non-linear solution can be
sensitive to the phasing of the individual time history. In order to keep the phases one applies to the signal a
real-only "transfer function" (i.e., with a zero-imaginary component), to rescale the Fourier amplitudes. The
technique is repeated iteratively until the desired matching is achieved for a certain range of periods. The more
iterations results with better compatibility with the target design spectrum (Ozdemir and Fahjan, 2007).

3.3. Spectral Matching in Time Domain

One approach for spectral matching is to adjust the original record iteratively in the time domain to achieve
compatibility with a specified target acceleration response spectrum by adding wavelets having specified period
ranges and limited durations to the input time history. These wave packets are added at times where there is
already significant amplitude in that period range in the time history. This method preserves the overall phasing
characteristics and as the time varying (i.e., non-stationary) frequency content of the ground motion
(Somerville, 1998). The resulting records each have an elastic response spectrum that is coincident (within a
tolerance) with the target spectrum. This procedure was first proposed by Kaul (1978) and was extended to
simultaneously match spectra at multiple damping values by Lilhanand and Tseng (1987). Although this
procedure is more complicated than the frequency domain matching procedure, in most cases it can preserves
the non-stationary character of the reference time history. Abrahamson (1992) developed RSPMATCH software
modifying the Lilhanand and Tseng algorithm that preserves non-stationary character of the reference ground
motion for a wider range of time histories.

3.4. Spectrum Compatible Artificial Record Generation

The principal goal of artificial accelerograms generation method is to obtain a design acceleration time history
that will have a response spectrum as close as desired to the predetermined target spectrum. Such accelerograms
can be obtained in frequency domain using different assumptions with iterative procedures. The computer
program SIMQKE (Gasparini and Vanmarcke, 1976) computes a power spectral density function from a
 th
The 14 World Conference on Earthquake Engineering
October 12-17, 2008, Beijing, China

specified smooth response spectrum and uses this function to derive the amplitudes of sinusoidal signals which
have random phase angles uniformly distributed between 0 and 2π. The sinusoidal motions are summed to
generate a time history record. An iterative procedure can be invoked to improve the match with the target
response spectrum, by calculating the ratio between the target and actual response ordinates at selected
frequencies. In order to get other characteristics of artificial spectrum-compatible record, such as duration, it is
necessary to obtain supplementary information about the expected earthquake motion apart from the response
spectrum. The code TARSCTHS (Papageorgiou et al., 2002) uses non-stationary stochastic vector processes to
generate artificial time histories from a user defined elastic response spectrum. The iterative scheme is applied
in frequency domain where the phase angles of the desired motion are randomly generated.

4. CRETERIA FOR EVALUATION OF MATCHED ACCELEROGRAMS

Even though, it is possible to obtain acceleration time histories that are almost completely compatible with the
elastic design spectrum by spectral matching or artificial acceleration generation, the resulting accelerograms
must have realistic characteristics of earthquakes in terms of amplitude, frequency content and duration. In
many cases, generated accelerograms have an excessive number of cycles of strong motion, and consequently
have unrealistically high energy content. Also, non-stationary characters of the initial time history such as,
availability P and S waves arrivals particularly at longer periods, can be altered if the shape of the Fourier
spectrum is changed significantly (Abrahamson, 1992). A baseline correction and filtering should be performed
to the generated acceleration time histories, and the corrected time-history should be checked again for
spectrum compatibility. The acceleration, velocity and displacement time histories should be examined to
ensure that they are reasonably close to the target values in terms of peak values, wave form, strong shaking
duration and other critical features such as the near-fault velocity pulse. Power spectral density function should
be examined to ensure a board distribution of energy in the final spectrum-compatible motion as a function of
Fourier period and there are no significant deficiencies in the energy at periods important to the structure.

5. EUROCODE 8 DESIGN SPECTRA and SELECTION CRITERIA for RECORDS

In Eurocode 8, the shape of horizontal elastic response spectrum is defined by four branches using the values of
the periods TB, TC and TD which take varying values for the five local site classes A, B, C, D, and E and two
ranges of earthquake magnitudes. It is recommended that the Type 2 spectrum is adopted for surface magnitude,
Ms, not greater than 5,5. Otherwise, Type 1 elastic response spectrum is used. Eurocode 8 permits using of
recorded, artificial and simulated accelerograms for earthquake time history analysis. A minimum of 3
accelerograms should be used where the mean of the zero period spectral response acceleration values should
not be smaller than the value of “ag S” for the site in question. The mean values of 5% damping elastic
spectrum, calculated from all time histories, should not be less than 90% of the corresponding values of the
code elastic response spectrum at periods range between 0,2T1 and 2T1, where T1 is the fundamental period of
the structure. For artificially generated acclereograms, when site-specific data are not available, the minimum
duration Ts of the stationary part of the accelerograms should not be less than 10 sec and the duration of the
accelerograms shall be consistent with the magnitude and the other relevant features of the seismic event
underlying the establishment of ag .

In this study, seven different design earthquakes have been generated by time scaling and SIMQKE,
TARSCTHS and RSPMatch programs to match Type 1 response spectrum. Considering both importance factor
and structural behavior factor to be unity, Type 1 design response spectra for ground type C and regional factor,
ag, to be 0.3g is selected as target response spectrum. The proposed design spectrum may represent the ground
motion of a point at epicentral distance of 15-30 km from an earthquake with moment magnitude of 6.6-7.5.
The candidate real accelerograms used for scaling and RSPMatch analyses are obtained from Pacific
Earthquake Engineering Research (PEER) Center, NGA strong motion database (PEER, 2005).
 th
The 14 World Conference on Earthquake Engineering
October 12-17, 2008, Beijing, China

6. DISCUSSIONS and CONCLUSIONS

The above mentioned four methods namely, Time scaling, SIMQKE, TARSCTHS and RSPmatch are used to
generate seven (7) acceleration time histories to fit the proposed design spectrum. Acceleration, velocity and
displacement spectra of the output time histories are compared with the correspondent target design spectra.
The nonlinear response of SDOF for the generated time histories are examined by computing the ductility
factor (µ) for different structural behavior factors, R=1,2,4,8. In real earthquake records, average of the ductility
factor is expected to be equal to structural behavior factor at longer periods (equal displacement rule) especially
for velocity and displacement sensitive spectral regions (Chopra, 2000). Equal displacement rule may not be
satisfied by time histories recorded near the source. In the Time scaling method the best fitted records with the
specified design spectrum are selected among the existing real time histories in Peer database and listed in
Table 1. The average of selected records acceleration, velocity and displacement spectra have some differences
up to %25 from the design spectra for some period ranges as it is noticed in Figure 1. The average nonlinear
response of the scaled time histories has compatible behavior with the equal displacement rule.

Table 1 List of earthquake records used for time scaling method

Figure 1 Acceleration, velocity and displacement spectra and ductility demand for time scaled records
 th
The 14 World Conference on Earthquake Engineering
October 12-17, 2008, Beijing, China

Figure 2 Acceleration, velocity and displacement spectra and ductility demand for SIMQKE generated records

Figure 3 Acceleration, velocity and displacement spectra and ductility demand for TARSCTH generated records
 th
The 14 World Conference on Earthquake Engineering
October 12-17, 2008, Beijing, China

Table 2 List of candidate earthquake records used for RSPMatch

Figure 4 Acceleration, velocity and displacement spectra and ductility demand for RSPMatch generated records

SIMQKE program is used to generate artificial acceleration time histories fitted with the proposed design
spectrum with the option to have trapezoidal intensity envelope with earthquake rise time, earthquake level
time, desired duration of acceleration and desired maximum ground acceleration as 2.5 sec, 12 sec, 35 sec and
0.3 g, respectively. The output time histories are filtered (0.1-20Hz) and corrected for the baseline. The average
of selected records acceleration, velocity and displacement spectra have inadequate fitting for longer periods
with differences up to %26 from the design spectra for some period ranges as it is noticed in Figure 2. The
average nonlinear response of the scaled time histories has incompatible behavior with the equal displacement
rule in velocity sensitive region. For option with no envelope, non realistic records that do not represent the
general characteristics of real earthquakes are generated. Even though the unfiltered time histories have the
same fit for the design spectrum as filtered time histories, unrealistic nonlinear behavior can be observed in
Figure 2 for the valid period range (0.1s-10s). TARSCTHS program requires moment magnitude and epicentral
 th
The 14 World Conference on Earthquake Engineering
October 12-17, 2008, Beijing, China

distance of scenario earthquake to compute duration. The same observation can be noticed for the acceleration
time histories artificially generated by TARSCTHS as shown in Figure 3. A better fitting to design spectrum but
worse estimation in the nonlinear response compared to SIMQKE can be noted.
RSPMatch program requires a set of earthquake records that initially have considerable compatibility with the
target spectrum. The list for initially selected records is given in Table 2. The matching procedure is not robust
and program requests a lot of parameters related to additional wavelet iterations. The convergence is not
guaranteed for many candidate records. The average of selected records acceleration, velocity and displacement
spectra have excellent fit with the design spectra for some period ranges as it is noticed in Figure 4. The average
nonlinear response of the scaled time histories has compatible behavior with the equal displacement rule.

REFERENCES

Abrahamson, N.A. (1992). Non-stationary spectral matching. Seismological Research Letters 63:1, 30-30.
Bommer, J.J., Acevedo, A.B. and Douglas, J. (2003). The selection and scaling of real earthquake accelerograms for use in
seismic design and assessment. Proceedings of ACI International Conference on Seismic Bridge Design and Retrofit,
American Concrete Institute.
Chopra AK (2000) Dynamics of Structures, 2nd Ed., Prentice Hall, New Jersey
Eurocode (2003). Design of structures for earthquake resistance. Part 1: General rules, seismic actions and rule for
buildings. Draft No. 49, October 2003, European Committee de Normalization, Brussels.
Fahjan Y.M., Ozdemir Z. and Al-Qaryouti, M. (2005). Selection and scaling of real earthquake accelerograms to fit the
new Jordanian design spectra, Paper No. 27, The International Earthquake Engineering Conference, November 21-24,
2005, Dead Sea/Jordan
Fahjan, Y.M., Ozdemir, Z. and Keypour, H. (2007). Procedures for real earthquake time histories scaling and application
to fit iranian design spectra, 5th International Conference on Seismology and Earthquake Engineering (SEE5), May 14-16,
2007, Tehran, Iran.
Gasparini, D.A. and Vanmarcke, E.H. (1976). Simulated earthquake motions compatible with prescribed response spectra,
Massachusetts Institute of Technology, Cambridge, Massachusetts.
Kaul, M.K. (1978). Spectrum consistent time-history generation, ASCE J. Eng Mech. EM4, 781-788.
Lilhanand, K. and Tseng, W.S. (1987). Generation of synthetic time histories compatible with multiple-damping response
spectra, SMIRT-9, Lausanne, K2-10.
Naeim, F. and Kelly, J.M. (1999). Design of Seismic Isolated Structures: From Theory to Practice, John Wiley & Sons.
Nikolaou, A.S. (1998). A GIS Platform for Earthquake Risk Analysis, Ph.D.Thesis, State University of New York at
Buffalo.
Ozdemir, Z. and Fahjan, Y.M. (2007). Comparison of time and frequency domain scaling of real accelerograms to match
earthquake design spectra, 6th National Conference on Earthquake Engineering (6. Ulusal Deprem Muhendisligi
Konferansi), October 16-20, 2007, Istanbul, Turkey.
Pacific Earthquake Engineering Research Center, 2005, PEER Strong Motion Database, http://peer.berkeley.edu/smcat/.
Papageorgiou, A., Halldorsson, B. and Dong, G. (2002). TARSCTH (Target Acceleration Spectra Compatible Time
Histories), Engineering Seismology Laboratory (ESL) at the State University of New York at Buffalo.
Preumont, A. (1984). The generation of spectrum compatible accelerograms for the design of nuclear power plants.
Earthquake Engineering and Structural Dynamics 12, 481-497.
Reiter, L. (1990) Earthquake Hazard Analysis: Issues and Insights, Columbia University Press, 254 pp.
Somerville, P. G. (1998). Emerging Art: Earthquake Ground Motion. Geotechnical Earthquake Engineerng and Soil
Dynamics III, ASCE Geotechnical Special Publication No. 75, Vol. 1, 1-38.
Stewart, J.P., Chiou, S.J., Bray, J.D., Graves, R.W., Somerville, P.G. and Abrahamson, N.A. (2001). Ground motion
evaluation procedures for performance-based design, PEER Report 2001/09, Pacific Earthquake Engineering Research
Center, University of California, Berkeley.